Why you can't trust stats on 'out-of-work benefit claims'
The current debate about welfare is being driven by some questionable numbers...
A new sense of a ‘welfare problem to be fixed’ is gaining ground. It’s one of the planks in the the current Government’s attempts to carve out a political dividing line on welfare, not least in Jeremy Hunt’s last Autumn Statement. But it’s also shared by Labour (at least partly), as well as nearly all journalists and think-tanks.
The claim is that we have a problem with a sharp rise in sick/disabled people not working and claiming benefits - particularly around mental health. This is seen as both a social problem and an unsustainable cost. And the commentariat mostly believe that this is a problem that needs to be tackled through reform of disability-related benefits, in both its incentives and the employment support it offers.
Yet I think many people are misunderstanding what is going here, partly because of some misleading numbers that are being bandied around, and partly in failing to understand what even accurate numbers mean. It’s not that this view is completely false, or that all of the policies being proposed are wrong-headed. Still, the misunderstandings are making people see the problem in the wrong light, missing some of the real issues that need fixing.
I’m going to work through this in a series of blog posts - starting here with the claim that out-of-work benefit claims have risen sharply.
Where do these numbers come from?
It’s rare to see anyone define ‘out-of-work benefits’, but as far as I can tell, this refers to working-age welfare benefits that are paid to people on the grounds that they are not working. (There’s various curiosities in who is included here, as I discuss below).
Several people that have claimed that out-of-work benefit claims are rising, including one by the Office for Budget Responsibility that we’ll come back to, and a spectacularly awful attempt by the editor of UnHerd. But I’m first going to focus on the version by Fraser Nelson in the Spectator, which seemed to get a lot of traction on Twitter/X, and is also admirably transparent about its data sources.1
Fraser’ Nelson’s chart is below - it seems to show that even after the effect of Covid-19 has ebbed, out-of-work benefit claims are now higher than they’ve been on record:
Fraser Nelson’s chart (not mine!):
It’s commendable that Fraser Nelson has put his sources here clearly, and if you check the sources, then you see these mostly come from official DWP series labelled ‘out-of-work benefits’ - which seems bona fide. Still, these series cannot be used to produce a trustworthy picture of how out-of-work benefits have changed over time. There have been some tweets debunking this, but I still don’t think it’s clear to many people why these claims are wrong (the issues are too complex for a tweet) - which is why I wanted to set this out as clearly as I could here.2
So what’s wrong with this claim?
The overarching problem is that Universal Credit (UC) has changed everything, so that the numbers post-UC mean something different to what they meant pre-UC (known as the ‘legacy system’). In particular, there are four big things that make these trends wholly untrustworthy. (This is a work in progress - so if you can help improve this post, please email me/add a comment below! )
Firstly, claimant’s partners are counted completely differently. Under UC, if you’re unemployed and in a couple (and have no other income), then both of you need to claim UC. Before UC, only one of you would usually be a claimant.3 (The best explanation of this is in the DWP Alternative Claimant Count methodology note - see the discussion of the Alternative Claimant Count in next week’s post). So a large part of why the numbers look larger on UC is that we’re counting extra people (out-of-work partners of claimants) when we’d previously not have included them in the count. I’m still astonished at this - it’s such a basic part of counting, and it’s so un-transparent in the data.
Secondly, some benefits are only counted as ‘out-of-work’ under UC, and not in the legacy system. In particular, out-of-work people claiming just Housing Benefit or Child Tax Credit are mostly not counted as ‘out-of-work claimants’ in the legacy system. (Again, this is clearly explained in the DWP Alternative Claimant Count methodology note). There’s also an issue with carers, as Carer’s Allowance is not counted as an out-of-work benefit (on the grounds that ‘they are not generally subject to labour market activation policies’), but someone on UC who gets the Carer’s Element is counted, and there’s also an increase in the share of CA claimants who also receive an out-of-work benefit, both of which skew the trend.
Third, people can now do more work and still be counted as an ‘out-of-work claimant’. This sounds a bit strange, but it’s because the definition of ‘out-of-work’ depends on whether people are earning above something called the Administrative Earnings Threshold (AET), and this has recently been raised. So for example, in early 2022 a couple working less than 14 hours between them (at the National Living Wage) was counted as ‘out-of-work’; in early 2023 this went up to 24 hours. As an aside, official statistics include an estimate of the number of unemployed claimants who are working, which is an interesting quirk!
Fourth, the size of the working-age population has changed - not just because of demographic shifts, but because the State Pension Age has been rising. Women aged 60-65 and men aged 65 are now ‘working-age’ when they weren’t in 2010 - and this isn’t always taken into account (or more commonly, it’s just not clear if it’s taken into account).
All of these will make the current ‘out-of-work claimants’ figure higher than it used to be. But there’s also a fifth issue where it’s difficult to know whether this will make the current figure higher or lower - some people are counted as ‘claimants’ even though they don’t receive any money. For legacy benefits, there are ‘credits-only’ JSA/ESA claims that are included in the main count, but where people don’t actually receive any money. (Many of these people will get Income Support instead, but not all of them). For UC, there are ‘nil claims’ where people are technically claimants but are not entitled to receive any money in a given month – which doesn’t apply in the legacy system (beyond marginal cases).
This isn’t an exhaustive list btw - but these are the issues I think matter most. But hopefully by setting this out, it becomes a bit clearer why the chart above - and similar comparisons - can’t be trusted.
But does any of this matter?
In short: this makes a big difference. All of these issues change the numbers by at least a hundred thousands people - and sometimes more. In next week’s post, I’ll try to put some numbers on them, and discuss whether it’s possible to get an out-of-work benefits trend that we can trust…
To be fair to Fraser Nelson, his claim is that (in the words of the article headline) “Yes, five million are on out-of-work benefits. Here’s the proof”. And this claim in itself is correct, as there are more than 5 million people on ‘out-of-work benefits’ as defined by DWP - it’s the trend over time that I’m critiquing here. Still, a lot of Fraser Nelson’s argument is about the recent trend (in the graph reproduced above, and in lots of the other charts/claims which are about changes since before Covid), so I don’t think he’s immune to the critique here.
Again, to be fair to Fraser Nelson, someone should have clearly explained this somewhere - indeed, his original post was an argument that the powers that be really should have put this trend together themselves.
This is despite the fact that there has long been a household means-test, as well as additions for adult dependents (or more recently, that cohabiting couples needed to make joint claims in the legacy benefits that existed just before UC was introduced). These joint claims are a bit of a puzzle - technically I think these were usually required from about 2000, but I’ve spoken to people who insist that they and their partner made separate claims, and it’s very hard to know how these are counted in the official data.